Data Cleaning - Geospatial Data

import pandas as pd
import os
import geopandas as gpd
import matplotlib.pyplot as plt
from geopy.geocoders import Nominatim
import folium
import ipywidgets as widgets
from IPython.display import display

# specify the directory containing the CSV files
directory = '../../data/quarterly_financials'

# create an empty list to store the dataframes
dfs = []

# loop over the CSV files in the directory
for filename in os.listdir(directory):
    if filename.endswith('.csv'):
        # read the CSV file into a dataframe and append it to the list
        path = os.path.join(directory, filename)
        df = pd.read_csv(path)
        dfs.append(df)

# concatenate the dataframes into a single dataframe
quarterly_financials = pd.concat(dfs, ignore_index=True)
# sort
quarterly_financials = quarterly_financials.sort_values(by=['REPDTE', 'CERT'], ascending=[False, True])
quarterly_financials = quarterly_financials.reset_index(drop=True)

# print the combined dataframe
display(quarterly_financials.shape)
display(quarterly_financials.head())

(1000105, 15)

	ZIP	BKCLASS	REPDTE	DEPDOM	ASSET	STNAME	EQ	NAME	CITY	ADDRESS	ENDEFYMD	CERT	ESTYMD	LIAB	ID
0	2111	SM	20221231	163284000.0	298020000	MASSACHUSETTS	26579000.0	STATE STREET BANK&TRUST CO	BOSTON	1 LINCOLN ST	99991231.0	14	17920101	271441000.0	14_20221231
1	36830	SM	20221231	952037.0	1023366	ALABAMA	65967.0	AUBURNBANK	AUBURN	100 N GAY ST	99991231.0	35	19070103	957399.0	35_20221231
2	36732	NM	20221231	407949.0	444822	ALABAMA	32577.0	ROBERTSON BANKING CO	DEMOPOLIS	216 N WALNUT AVE	99991231.0	39	18700101	412245.0	39_20221231
3	36867	NM	20221231	266874.0	265272	ALABAMA	-8465.0	PHENIX-GIRARD BANK	PHENIX CITY	801 13TH ST	99991231.0	41	19040504	273737.0	41_20221231
4	36401	NM	20221231	70649.0	76239	ALABAMA	5487.0	BANK OF EVERGREEN	EVERGREEN	146 W FRONT ST	99991231.0	49	19320901	70752.0	49_20221231

display(len(quarterly_financials['BKCLASS'].unique()))
display(quarterly_financials['BKCLASS'].unique())

7

array(['SM', 'NM', 'N', 'SI', 'SB', 'SL', 'OI'], dtype=object)

Zip coordinates

# bring in
zip_coordinates = pd.read_csv('../../data/coordinate_data/2022_Gaz_zcta_national.txt', sep='\t')

# remove whitespace in col names
zip_coordinates.columns = [col.strip() for col in zip_coordinates.columns]

# drop unnecesary columns
zip_coordinates.drop(['ALAND', 'ALAND_SQMI', 'AWATER', 'AWATER_SQMI'], axis=1, inplace=True)

# check
display(zip_coordinates.shape)
display(zip_coordinates.head())

(33791, 3)

	GEOID	INTPTLAT	INTPTLONG
0	601	18.180555	-66.749961
1	602	18.361945	-67.175597
2	603	18.457399	-67.124867
3	606	18.158327	-66.932928
4	610	18.293960	-67.127182

display(len(quarterly_financials['ZIP'].unique()))

10756

Combine

# Rename the 'GEOID' column in 'zip_coordinates' to match the 'ZIP' column in 'quarterly_financials'
zip_coordinates = zip_coordinates.rename(columns={'GEOID': 'ZIP'})

# Merge the two dataframes on the 'ZIP' column
merged_df = quarterly_financials.merge(zip_coordinates, on='ZIP', how='left')

# Create a new column 'coordinates' with the combined 'INTPTLAT' and 'INTPTLONG' columns as a tuple
merged_df['zip_coordinates'] = list(zip(merged_df['INTPTLAT'], merged_df['INTPTLONG']))

# Update the 'quarterly_financials' dataframe with the new 'coordinates' column
quarterly_financials = merged_df

# Print the updated 'quarterly_financials' dataframe
display(quarterly_financials.shape)
display(quarterly_financials.head())

(1000105, 18)

	ZIP	BKCLASS	REPDTE	DEPDOM	ASSET	STNAME	EQ	NAME	CITY	ADDRESS	ENDEFYMD	CERT	ESTYMD	LIAB	ID	INTPTLAT	INTPTLONG	zip_coordinates
0	2111	SM	20221231	163284000.0	298020000	MASSACHUSETTS	26579000.0	STATE STREET BANK&TRUST CO	BOSTON	1 LINCOLN ST	99991231.0	14	17920101	271441000.0	14_20221231	42.350680	-71.060527	(42.35068, -71.060527)
1	36830	SM	20221231	952037.0	1023366	ALABAMA	65967.0	AUBURNBANK	AUBURN	100 N GAY ST	99991231.0	35	19070103	957399.0	35_20221231	32.534872	-85.493755	(32.534872, -85.493755)
2	36732	NM	20221231	407949.0	444822	ALABAMA	32577.0	ROBERTSON BANKING CO	DEMOPOLIS	216 N WALNUT AVE	99991231.0	39	18700101	412245.0	39_20221231	32.417456	-87.892213	(32.417456, -87.892213)
3	36867	NM	20221231	266874.0	265272	ALABAMA	-8465.0	PHENIX-GIRARD BANK	PHENIX CITY	801 13TH ST	99991231.0	41	19040504	273737.0	41_20221231	32.498054	-85.023590	(32.498054, -85.02359)
4	36401	NM	20221231	70649.0	76239	ALABAMA	5487.0	BANK OF EVERGREEN	EVERGREEN	146 W FRONT ST	99991231.0	49	19320901	70752.0	49_20221231	31.468970	-86.950426	(31.46897, -86.950426)

# removed unmatched zips
quarterly_financials.dropna(subset=['INTPTLAT'], inplace=True)

# drop unnecesary columns
quarterly_financials.drop(['INTPTLAT', 'INTPTLONG', 'ENDEFYMD', 'CERT', 'ESTYMD', 'LIAB', 'ID', 'ADDRESS', 'EQ', 'ZIP'], axis=1, inplace=True)

# convert numerical values to millions (source is in thousands)
quarterly_financials['DEPDOM'] = quarterly_financials['DEPDOM'].divide(1000)
quarterly_financials['ASSET'] = quarterly_financials['ASSET'].divide(1000)

# rename
quarterly_financials.rename(columns={
    'BKCLASS': 'bank_class', 
    'REPDTE': 'report_date', 
    'DEPDOM': 'deposits_mill', 
    'ASSET': 'assets_mill', 
    'STNAME': 'state', 
    'NAME': 'name',
    'CITY': 'city'
    }, inplace=True)

display(quarterly_financials.head())

	bank_class	report_date	deposits_mill	assets_mill	state	name	city	zip_coordinates
0	SM	20221231	163284.000	298020.000	MASSACHUSETTS	STATE STREET BANK&TRUST CO	BOSTON	(42.35068, -71.060527)
1	SM	20221231	952.037	1023.366	ALABAMA	AUBURNBANK	AUBURN	(32.534872, -85.493755)
2	NM	20221231	407.949	444.822	ALABAMA	ROBERTSON BANKING CO	DEMOPOLIS	(32.417456, -87.892213)
3	NM	20221231	266.874	265.272	ALABAMA	PHENIX-GIRARD BANK	PHENIX CITY	(32.498054, -85.02359)
4	NM	20221231	70.649	76.239	ALABAMA	BANK OF EVERGREEN	EVERGREEN	(31.46897, -86.950426)

# change to title case
quarterly_financials[['state', 'name', 'city']] = quarterly_financials[['state', 'name', 'city']].apply(lambda x: x.str.title())

# format date
quarterly_financials['report_date'] = pd.to_datetime(quarterly_financials['report_date'], format='%Y%m%d')

# set to float
quarterly_financials['deposits_mill'] = quarterly_financials['deposits_mill'].astype(float)
quarterly_financials['assets_mill'] = quarterly_financials['assets_mill'].astype(float)


display(quarterly_financials.head())

	bank_class	report_date	deposits_mill	assets_mill	state	name	city	zip_coordinates
0	SM	2022-12-31	163284.000	298020.000	Massachusetts	State Street Bank&Trust Co	Boston	(42.35068, -71.060527)
1	SM	2022-12-31	952.037	1023.366	Alabama	Auburnbank	Auburn	(32.534872, -85.493755)
2	NM	2022-12-31	407.949	444.822	Alabama	Robertson Banking Co	Demopolis	(32.417456, -87.892213)
3	NM	2022-12-31	266.874	265.272	Alabama	Phenix-Girard Bank	Phenix City	(32.498054, -85.02359)
4	NM	2022-12-31	70.649	76.239	Alabama	Bank Of Evergreen	Evergreen	(31.46897, -86.950426)

quarterly_financials['bank_class'] = quarterly_financials['bank_class'].replace({
    'N':  'Commercial bank, national charter, Fed member',
    'NM': 'Commercial bank, state charter, Fed non-member',
    'OI': 'Insured U.S. branch of a foreign chartered institution',
    'SB': 'Federal savings banks',
    'SI': 'State chartered stock savings banks',
    'SL': 'State chartered stock savings and loan association',
    'SM': 'Commercial bank, state charter, Fed member',
    'NC': 'Noninsured non-deposit commercial bank',
    'NS': 'Noninsured stock savings bank',
    'CU': 'State or federally chartered credit union',
    })

display(quarterly_financials.shape)
display(quarterly_financials.head())

(965390, 8)

	bank_class	report_date	deposits_mill	assets_mill	state	name	city	zip_coordinates
0	Commercial bank, state charter, Fed member	2022-12-31	163284.000	298020.000	Massachusetts	State Street Bank&Trust Co	Boston	(42.35068, -71.060527)
1	Commercial bank, state charter, Fed member	2022-12-31	952.037	1023.366	Alabama	Auburnbank	Auburn	(32.534872, -85.493755)
2	Commercial bank, state charter, Fed non-member	2022-12-31	407.949	444.822	Alabama	Robertson Banking Co	Demopolis	(32.417456, -87.892213)
3	Commercial bank, state charter, Fed non-member	2022-12-31	266.874	265.272	Alabama	Phenix-Girard Bank	Phenix City	(32.498054, -85.02359)
4	Commercial bank, state charter, Fed non-member	2022-12-31	70.649	76.239	Alabama	Bank Of Evergreen	Evergreen	(31.46897, -86.950426)

quarterly_financials['zip_coordinates'].isna().sum()

0

Growth of Assets Per State

from IPython.display import display, HTML
from io import BytesIO
import base64
import os
import pandas as pd
import geopandas as gpd
import folium


# Convert quarterly_financials to a GeoDataFrame
gdf = gpd.GeoDataFrame(quarterly_financials, geometry=gpd.points_from_xy(quarterly_financials.zip_coordinates.apply(lambda p: p[1]), quarterly_financials.zip_coordinates.apply(lambda p: p[0])))

def plot_bank_assets_by_date(date):
    # Filter the GeoDataFrame by the selected date
    filtered_gdf = gdf[gdf['report_date'] == date]

    # Group the filtered GeoDataFrame by state and sum the assets
    state_assets = filtered_gdf.groupby('state')['assets_mill'].sum().reset_index()

    # Create the base folium map
    m = folium.Map(location=[37.8, -96], zoom_start=3)

    # Define a function to scale the assets to a suitable size for the map
    def scale_bubble_size(assets):
        return assets / 50000

    # Plot bubbles for each state with a size proportional to the total assets
    for index, row in state_assets.iterrows():
        state_data = filtered_gdf[filtered_gdf['state'] == row['state']]
        state_centroid = state_data.unary_union.centroid
        folium.CircleMarker(
            location=[state_centroid.y, state_centroid.x],
            radius=scale_bubble_size(row['assets_mill']),
            color='blue',
            fill=True,
            fill_color='blue',
            fill_opacity=0.5,
            popup=f"State: {row['state']}<br>Total Assets: {row['assets_mill']:.0f}<br>Date: {date}"
        ).add_to(m)

    # Display the map
    return m


unique_dates = ['1992-12-31', '2002-12-31', '2012-12-31', '2022-12-31']

maps = []

for date in unique_dates:
    maps.append(plot_bank_assets_by_date(pd.to_datetime(date)))

    
# for rendering in HTML
def folium_map_to_iframe_data(m, map_name):
    map_path = f'geospatial_data_html/{map_name}.html'
    m.save(map_path)
    iframe = IFrame(src=map_path, width=350, height=300)
    return iframe

# Initialize the HTML structure for the 2-column layout
html_structure = """
<style>
    .map-container {
        display: flex;
        flex-wrap: wrap;
    }
    .map-box {
        width: 50%;
        padding: 5px;
        box-sizing: border-box;
    }
    .map-title {
        text-align: center;
        font-weight: bold;
    }
</style>
<div class="map-container">
"""

# Add the iframes to the HTML structure with the date on top
for i, m in enumerate(maps):
    map_name = f'map_{i}'
    iframe_data = folium_map_to_iframe_data(m, map_name)
    date = unique_dates[i]
    html_structure += f'<div class="map-box"><div class="map-title">{date}</div>{iframe_data._repr_html_()}</div>'

# Close the map-container div and display the HTML structure
html_structure += '</div>'
display(HTML(html_structure))



# For rendering in Jupyter for verification
if True == False:

    def folium_map_to_iframe_data(m):
        data = BytesIO()
        m.save(data, close_file=False)
        data.seek(0)
        b64 = base64.b64encode(data.read()).decode("utf-8")
        return f'<iframe src="data:text/html;base64,{b64}" width="350" height="300"></iframe>'

    table_html = '<table><tr>'
    for i, m in enumerate(maps):
        iframe_data = folium_map_to_iframe_data(m)
        table_html += f'<td><h3>{unique_dates[i]}</h3>{iframe_data}</td>'
        if (i + 1) % 2 == 0:
            table_html += '</tr><tr>'
    table_html += '</tr></table>'

    display(HTML(table_html))

1992-12-31

2002-12-31

2012-12-31

2022-12-31

Individual Points

Heatmap all dates

Latest date only

import folium
from folium.plugins import HeatMap
import pandas as pd
from IPython.display import IFrame

# Find the latest date
latest_date = quarterly_financials['report_date'].max()

# Filter the DataFrame by the latest date
latest_quarterly_financials = quarterly_financials[quarterly_financials['report_date'] == latest_date]

# Create a base map
map = folium.Map(location=[37.8, -96], zoom_start=4)

# Prepare the data for the HeatMap
heatmap_data = []
for index, row in latest_quarterly_financials.iterrows():
    coordinates = row['zip_coordinates']
    assets_mill = row['assets_mill']
    heatmap_data.append((*coordinates, assets_mill))

# Add the HeatMap to the map
heatmap = HeatMap(heatmap_data, radius=13, max_zoom=13, gradient={0.2: 'blue', 0.4: 'lime', 0.6: 'orange', 1: 'red'})
map.add_child(heatmap)

# Function to scale the assets to a suitable size for the map
def scale_bubble_size(assets):
    return assets / 50000

# Add CircleMarkers with tooltips to the map
for index, row in latest_quarterly_financials.iterrows():
    coordinates = row['zip_coordinates']
    assets_mill = row['assets_mill']
    bank_name = row['name']
    tooltip_text = f"Bank: {bank_name}<br>Coordinates: {coordinates}<br>Assets (millions): {assets_mill}"
    folium.CircleMarker(
        location=coordinates,
        radius=scale_bubble_size(assets_mill),
        color='blue',
        fill=True,
        fill_color='blue',
        fill_opacity=0.5,
        tooltip=tooltip_text,
    ).add_to(map)

# For rendering on Quarto
map.save('geospatial_data_html/heatmap.html')
IFrame(src='geospatial_data_html/heatmap.html', width=700, height=600)

# For rendering in Jupyter Notebook
display(map)

Make this Notebook Trusted to load map: File -> Trust Notebook